Provenance of Oligocene–Miocene sedimentary rocks in the Cuu Long and Nam Con Son basins, Vietnam and early history of the Mekong River

The offshore Cuu Long (CLB) and Nam Con Son (NCSB) basins of SE Vietnam are two important Cenozoic hydrocarbon-bearing basins in the southern South China Sea (SCS), which can contribute to understanding the evolution of major SE Asian river systems, in particular the Mekong River. The Oligocene to Early Miocene basin fill of the Cuu Long Basin is dominated by sediment sourced locally from the Da Lat Zone basement on land. Sandstones have abundant Cretaceous detrital zircons and heavy mineral assemblages dominated by apatite and epidote. The Bach Ho Unconformity at c. 16 to 20.5 Ma marks a major provenance change and the overlying Middle to Late Miocene Con Son and Dong Nai formations were sourced by a large river system, which drained Indochina or even the Himalaya, resembling the present-day Mekong River. These formations have heterogeneous detrital zircon populations dominated by Triassic ages and zircon–rutile–tourmaline-dominated heavy mineral assemblages. The Oligocene Cau and Early Miocene Dua formations of the Nam Con Son Basin have a similar provenance to the CLB Con Son and Dong Nai formations, indicating a comparable drainage history of a large proto-Chao Phraya. At the Dua Unconformity at c. 15.4 to 17 Ma the Indochina provenance signature changes to a predominant Sibumasu signature. The early Mekong River evolution is suggested to have involved two river systems. A proto-Mekong 1 initially filled the CLB, while at the same time a proto-Chao Phraya River with a capture area into the Himalaya filled the Malay Basin and/or the NCSB. At the end of the Early Miocene drainage was completely reorganised, and the proto-Mekong 1 captured the headwaters of the proto-Chao Phraya and became the proto-Mekong 2, while the NCSB was filled by sediment from the Malay–Thai Peninsula. This major change marks a very important drainage reorganisation in the southern SCS region.


Introduction
The Cenozoic basins of the South China Sea (SCS) margin have a long history of successful hydrocarbon exploration and production. Opening of the South China Sea in the Cenozoic, as well as sea level changes and hinterland uplift resulted in significant changes in depositional environments, input of clastic material and drainage into the basins. In the offshore region of SE Vietnam (Fig. 1) are the prolific Cuu Long (CLB) and Nam Con Son (NCSB) basins, which are the subject of this sediment provenance study.
The Yinggehai-Song Hong Basin in the Gulf of Tonkin ( Fig. 1) has so far provided the most detailed provenance data from offshore Vietnam (e.g., Wang et al. 2014Wang et al. , 2019Cao et al. 2015;Jiang et al. 2015;Fyhn et al. 2019). However, up to now the offshore region of SE Vietnam has been little studied due to limited sample availability, with the exception of preliminary provenance data from the Cuu Long Basin (Hennig et al. 2017a).
The Cuu Long and Nam Con Son basins are close to the SE Vietnam coast and the present-day Mekong River supplies sediment to the southern Cuu Long Basin (Fig. 1). The Mekong is the sixth longest river in Asia (Liu et al. 2009) and is very important for agriculture, climate and human development. Provenance studies of Cenozoic successions can help to understand the evolution of the Mekong River system by identifying the history of sediment deposition, uplift of source regions, changes of river courses, and provide insights into causes, such as tectonics and climate change. The objective of this study was to investigate the provenance of Oligocene to Late Miocene clastic sedimentary rocks offshore SE Vietnam to reconstruct the drainage history of major rivers in Indochina. We present here new heavy mineral data and detrital zircon geochronology based on samples from wells in the Cuu Long and Nam Con  Clift et al. 2006;). The study area includes the Cuu Long and Nam Con Son basins (orange coloured) offshore SE Vietnam. Other important SCS basins coloured in yellow. The Da Lat Zone in SE Vietnam (Cretaceous granitoids) is coloured in dark grey as important proximal source region 1 3 Son basins to analyse river course and provenance changes. These are the first provenance data from these two basins. The results demonstrate that there is a major provenance change in the Cuu Long Basin in the Early Miocene from a proximal Da Lat Zone source to a wider Indochina source, which we interpret as reflecting the evolution of the proto-Mekong River. At around the same time in the Nam Con Son Basin there was a change from an Indochina to an Indochina-Sibumasu source, indicating a change in hinterland supply and drainage.

Regional background
The Sunda Shelf region includes several Cenozoic sedimentary basins that began to extend in the Late Eocene (Hall and Morley 2004). Rifting along the Sunda Shelf mostly ceased or paused when oceanic spreading began in the South China Sea at the Eocene-Oligocene boundary (Barckhausen et al. 2014;Larsen et al. 2018;Zhong et al. 2018), which continued until c. 15 or 16 Ma (Chang et al. 2015;Yu et al. 2018;Zhang et al. 2018;Zhong et al. 2018). Franke et al. (2014) identified two important unconformities in the South China Sea margins, the rift-onset unconformity and the end-rift or break-up unconformity with diachronous character and Paleocene to Eocene and Oligocene ages, respectively. The Cuu Long and Nam Con Son basins are two NE-SW trending basins in offshore SE Vietnam that are separated by the Con Son Swell. The basins contain Late Eocene/Oligocene to Miocene sediments that were deposited in fluvial to marine environments, which have been documented by 2D/3D seismic, wireline logs, biostratigraphy, palynology, and core petrography (e.g., Nguyen and Nguyen 1995;Matthews et al. 1997;Dien et al. 1998;Lee et al. 2001;Morley et al. 2011Morley et al. , 2019Dung et al. 2018). These sediments were deposited above Mesozoic meta-sedimentary, metamorphic and igneous rocks that form the basement (e.g., Matthews et al. 1997;Cuong and Warren 2009). Figure 2 displays the Cenozoic stratigraphy of onshore and offshore SE Vietnam. Despite their close spatial relationship and similar sediment fill, the basins differ in their structure and evolution (e.g., Lee et al. 2001;Schmidt et al. 2019). Onshore, the Da Lat Zone in SE Vietnam consists predominantly of Cretaceous granitoids intruded into Triassic and Jurassic meta-sedimentary and Fig. 2 Stratigraphic overview of the offshore successions in the Cuu Long and Nam Con Son basins and their time equivalents in onshore southern Vietnam (modified from . Letters in the CLB and NCSB represent the seismic horizons Triassic igneous rocks (Nguyen et al. 2004;Fyhn et al. 2010;Shellnutt et al. 2013;Hennig-Breitfeld et al. 2021;Schmidt et al. 2021). This Cretaceous granitic belt extends into SW Vietnam and SE Cambodia (Fyhn et al. 2016;Nong et al. 2021;Waight et al. 2021) and across the Sunda Shelf into Borneo (Hennig et al. 2017b;Breitfeld et al. 2017Breitfeld et al. , 2020b. The Cretaceous granites and volcanics in the Da Lat Zone are overlain by mainly fluvial to marginal marine Cenozoic sediments (Fig. 2).  presented heavy mineral and detrital zircon age provenance data for the Di Linh, Ba Mieu and Song Luy formations in the Da Lat Zone, which were probably largely eroded in the Pleistocene to Quaternary.
The Cuu Long Basin formed in the Cenozoic and is filled with c. 6 to 10 km of sediments (Lee et al. 2001;Cuong and Warren 2009;Fyhn et al. 2009;Schmidt et al. 2019) and has a history which includes rifting, inversion/compression, and thermal subsidence. Rifting initiated in the Eocene and continued until the end of the Oligocene (Lee et al. 2001;Schmidt et al. 2003;Hung and Le 2004;Fyhn et al. 2009) or the Early Miocene (Cuong and Warren 2009). Nguyen et al. (2019) documented two rift periods, a main phase from the Middle Eocene to the Oligocene and a late phase in the Early Miocene. This was followed by an episode of compression in the late Oligocene to Early Miocene (Schmidt et al. 2003(Schmidt et al. , 2019Hung and Le 2004;Cuong and Warren 2009;Fyhn et al. 2009). Schmidt et al. (2019) suggested rifting ceased in the mid-Oligocene, followed by inversion in the late Oligocene. The last stage of basin evolution is interpreted as a thermal sag phase beginning in the Early Miocene (Lee et al. 2001;Fyhn et al. 2009) or Middle Miocene (Cuong and Warren 2009;Nguyen et al. 2019). Schmidt et al. (2019) noted that none of the faults in the Mekong Delta Fault Zone or in the Cuu Long Basin cut significantly into the adjacent Con Son Swell or into the Nam Con Son Basin.
The Nam Con Son Basin is a Cenozoic rift basin that has been interpreted to be directly associated with the rifting and seafloor spreading of the South China Sea (East Vietnam Sea in Vietnamese literature) (Briais et al. 1993;Fyhn et al. 2009;Franke et al. 2014), although the South China Sea extension estimates of Hayes and Nissen (2005) cast doubt on how much influence spreading had on the SE Vietnam basins. Rifting of the Nam Con Son Basin started at c. 32 Ma (Swiecicki and Maynard 2009), later than many other basins across the Sunda region, where an earlier Eocene rift phase is commonly identified (Morley et al. 2011). Four principal megasequences based on tectonic stages of the Nam Con Son Basin have been identified and include syn-rift 1, interrift or post-rift 1, syn-rift 2 and post-rift 2 (Matthews et al. 1997;Lee et al. 2001;Morley et al. 2011;Meyer 2012;Tuan and Tri 2016;Dung et al. 2018). The megasequences are further subdivided into sequence stratigraphic third-order depositional sequences (T10 to Q100) that are correlated with lithostratigraphic units. The syn-rift 1 consists of sequence T10 deposited in the early Oligocene (Morley et al. 2011), possibly ranging down into the Late Eocene (Matthews et al. 1997;Dung et al. 2018). Inter-rift consists of T20 and T30 associated with regional subsidence possible in response to opening of the South China Sea. Syn-rift phase 2 includes sequences T40, T50, T60 and T65. This phase was followed by a regional uplift event in the Middle Miocene associated with major erosion, resulting in a regional Middle Miocene Unconformity (Matthews et al. 1997;Lee et al. 2001;Fyhn et al. 2009;Swiecicki and Maynard 2009;Morley et al. 2011;Tuan and Tri 2016;Dung et al. 2018). Following the inversion widespread subsidence resulted in deposition of post-rift 2 sequences T85 to Q100, which includes Plio-Pleistocene delta sediments that were interpreted as the prograding paleo-Mekong delta (Matthews et al. 1997;Morley et al. 2011;Tuan and Tri 2016).

Cuu Long Basin stratigraphy
The stratigraphy of the Cuu Long Basin is subdivided by seismic horizons labelled from Group F/G at the base to A at the top (e.g., Dien et al. 1998;Hung and Le 2004) (Fig. 3). The basement of the basin is predominantly formed by Cretaceous granites, with minor Late Triassic and Late Jurassic igneous rocks (Cuong and Warren 2009). Hennig-Breitfeld et al. (2021) reported a U-Pb zircon age of c. 90-92 Ma for the Cretaceous granite basement and concluded that it was the equivalent of the onshore Da Lat Zone Deo Ca Suite granites. The Bach Ho granite within the basin was dated by Nong et al. (2021) with zircon U-Pb as c. 110 Ma, more similar in age to the onshore Dinh Quan suite (Hennig-Breitfeld et al. 2021). The Con Dao Island granite southwest of the Cuu Long Basin was dated by Nong et al. (2021) as c. 83 Ma, similar to the onshore Ankroet suite (Hennig-Breitfeld et al. 2021). This suggests that the Cretaceous basement of the Cuu Long Basin is formed by different Cretaceous granitoids similar to the onshore Da Lat Zone. The sedimentary successions of the Cuu Long Basin begin locally with the Eocene Ca Coi Formation (seismic groups F and G) and the more widespread Late Eocene to early Oligocene Tra Cu Formation (seismic group E) (Hung and Le 2004;San et al. 2013;Quang 2013;Morley et al. 2019). These formations consist of alluvial, fluvial and lacustrine fine to coarse grained sandstones interbedded with conglomerates, and siltstones, minor mudstones and thin coals (Cuong and Warren 2009;Morley et al. 2019). They are unconformably overlain by the Tra Tan Formation (seismic groups C and D) (Morley et al. 2019). Depositional environments range from fluvial and lacustrine to brackish and intertidal with some interbedded volcaniclastics (Hung and Le 2004;Cuong and Warren 2009;San et al. 2013;Morley et al. 2019). An early Oligocene age is interpreted for the formation by Morley et al. (2019) in contrast to the previous Oligocene or late Oligocene interpretation (e.g., Cuong and Warren 2009). The formation is subdivided into Lower (seismic group D) and Upper (seismic group C) parts separated by an unconformity (Morley et al. 2019). The Bach Ho Formation (seismic group BI.1 and BI.2) unconformably overlies the Tra Tan Formation and consists mainly of shallow marine shales and sands, and some mudstones with strong deltaic and fluvial influence (Cuong and Warren 2009;Morley et al. 2019

Nam Con Son Basin stratigraphy
The Nam Con Son Basin is subdivided into seismic sequences labelled from T10 (Late Eocene to early Oligocene) to Q100 (Pleistocene) corresponding to various lithostratigraphic units ( Fig. 4) (Matthews et al. 1997). Based on the seismic sequences, Morley et al. (2011) introduced sequence biostratigraphy cycles VIM (Vietnam/Indonesia/ Malaysia) and revised the ages for the corresponding formations. Therefore, the seismic sequences of Matthews et al. (1997) do not correspond exactly to the lithostratigraphic formations (Morley et al. 2011). The basement is formed by granitic, volcanic and metasedimentary rocks (Matthews et al. 1997), from which volcanic rocks have been dated by zircon U-Pb as 200.4 ± 1.6 Ma (Hennig-Breitfeld et al. 2021). Cretaceous granitoids have also been reported by the K-Ar method (Hutchison 1989;Areshev et al. 1992). Above the basement T10 and VIM10 forms the first sedimentary package consisting of alluvial fan/fluvial to lacustrine deposits, and is assumed to be Late Eocene to early Oligocene (Matthews et al. 1997;Morley et al. 2011;Dung et al. 2018). The T10 sequence has not been penetrated anywhere in the basin, and its age must be inferred from the overlying T20 sequence. The T10 seismic sequence corresponds to the Tien Cau Formation (Morley et al. 2011). T20 and VIM20 is dominated by fluvial sandstones overlain by coals, mudstones and thinner sandstone packages and interpreted as backstepping fluvio-deltaic facies with minimal tidal influence (Matthews et al. 1997;Morley et al. 2011). VIM20 corresponds to the Cau Formation and is dated as late Oligocene (Bat et al. 1993;Matthews et al. 1997). Nguyen and Nguyen (1995) report a thickness of 200 to 800 m for the Cau Formation. T30 and VIM30 unconformably above are interpreted as late Oligocene to Early Miocene transgressive sandstones and marine mudstones overlain by interbedded sandstones and mudstones (Matthews et al. 1997;Morley et al. 2011;Dung et al. 2018). They are fluvial, estuarine and brackish shoreface deposits with occasional intercalations of more open marine sediments (Matthews et al. 1997). T30 and VIM30 correspond to the Dua Formation (Bat et al. 1993;Nguyen and Nguyen 1995;Matthews et al. 1997;Morley et al. 2011). Nguyen and Nguyen (1995) report a thickness from 200 m to over 1000 m. Tham and Su (2016) reported Early Miocene calcareous nannofossils (zones NN2 to NN4). Seismic sequence T40 and VIM40 is interpreted to be unconformably above with a shift from the predominantly paralic facies of T30 to shelfal marine clastic sediments and carbonates of T40 (Matthews et al. 1997). The sequences correspond to the Thong Formation and are dated as late Early Miocene to early Middle Miocene (Bat et al. 1993;Matthews et al. 1997;Morley et al. 2011;Dung et al. 2018). The Thong Formation also includes seismic sequence T50 (Dung et al. 2018). The Middle Miocene T60 and T65 are dominated by shelfal sediments and deep water clastic sediments (Matthews et al. 1997). They correspond to the Mang Cau Formation (Bat et al. 1993;Nguyen and Nguyen 1995;Matthews et al. 1997;Dung et al. 2018), and to cycle VIM50 (Morley et al. 2011). The top of the sequence is a prominent unconformity marking a basinward shift of deep water sedimentation and mild inversion of the underlying successions (Matthews et al. 1997). T80 and T85 are unconformably above T65 and comprise Late Miocene deposits (Matthews et al. 1997). They correspond broadly with the VIM70 cycle (Morley et al. 2011). The lower part (T80) is dominated by widespread shelf carbonates with restricted shallow water clastics and deep water mudstones between the build-ups (Matthews et al. 1997). The upper part (T85) is composed of shelf and slope clastic deposits with some restricted carbonate facies, and turbidite facies (Matthews et al. 1997). The Nam Con Son Formation is the lithostratigraphic equivalent (Bat et al. 1993;Matthews et al. 1997). Sequences T90, T100 and Q100 correspond to the Bien Dong Formation and broadly to VIM cycles 80, 90 and 100, and range from the Early Pliocene to the Pleistocene with mostly shelfal, slope and deep water deposits (Bat et al. 1993;Matthews et al. 1997;Morley et al. 2011;Dung et al. 2018) and delta deposits (Morley et al. 2011).

Sampling
Samples were provided from wells in the Cuu Long and Nam Con Son basins by Talisman Vietnam (now Repsol Vietnam) in cooperation with the Vietnam Petroleum Institute (VPI). One well from the Cuu Long Basin was analysed using six cuttings samples from various stratigraphic levels (Fig. 3). The Nam Con Son Basin was analysed in three wells (NCSa, NCSb, NCSc) covering Oligocene to Miocene stratigraphic intervals (Fig. 4). Three cuttings samples Fig. 4 Stratigraphy of the Nam Con Son Basin with depositional environments, tectonic phases, and provenance signature (modified from Dung et al. 2018). Seismic sequences T20 to T40&50 (Cau to Thong Formation) were analysed in this study. Display of the three analysed wells (NCSa, NCSb and NCSc) and position of the samples is only schematically. Red dots include heavy mineral and detrital zircon geochronology samples. Grey dots comprise biostratigraphy samples ◂ were analysed from well NCSa, five cuttings samples from well NCSb, and two core samples from NCSc. A sample list overview with analysis undertaken can be found in Supplementary Table 1.

Biostratigraphy
Covered thin sections were prepared for calcareous mudstones of 16 cuttings samples of the Nam Con Son Basin (wells NCSa and NCSb) and two cuttings samples of the Cuu Long Basin, and analysed for biostratigraphy, following the approach described in BouDagher-Fadel (2015, 2018a. This primarily uses the Planktonic Zonation scheme (PZ) of BouDagher-Fadel (2018b), which is tied to the biostratigraphical and the radioisotope time scales (as defined by Gradstein et al. 2012 and revised by Cohen et al. 2013).

Heavy mineral separation and analysis
Sample preparation for heavy mineral analyses and zircon enrichment was carried out at Royal Holloway University of London. The friable cuttings and core samples were processed by mortar and pestle, washed and sieved into a 63-250 μm fraction. Heavy minerals were separated using standard heavy liquid lithium heteropolytungstate at a density of 2.89 g/cm 3 , using the technique outlined in Mange and Maurer (1992). Heavy mineral analysis was performed primarily on polished epoxy resin blocks with QEMSCAN by Rocktype with QEMSCAN 10 measurement method that has a resolution of 10 μm. Additional samples from the Cuu Long Basin were analysed with EDS-SEM at the Royal Holloway University of London after the methodology outlined in Hennig-Breitfeld et al. (2019) and Breitfeld et al. (2020a), and Raman spectroscopy (e.g., Andò et al. 2009;Andò and Garzanti, 2014) was conducted on samples from both basins at Chemostrat Ltd. on a Horiba XploRa Raman with a 532 µm laser. Acquired spectra were compared to the RRUFF database (Lafuente et al. 2016) and an internal Chemostrat Ltd. database. Dunkl et al. (2020) demonstrated that electron beambased methods such as QEMSCAN yield overall results that are comparable to Raman analysis, especially if the heavy mineral assemblage does not contain high proportions of polymorphs (which are not resolvable with electron beam methods) and chain silicates (which are difficult to discriminate with electron beam methods). Supplementary Table 1 displays which method was used for each sample. QEM-SCAN results yielded the highest count numbers.

Baryte removal technique
A few cuttings samples contained abundant heavy minerals baryte or anhydrite. Baryte is indicative of drilling mud contamination and anhydrite may also be a drilling additive or be diagenetic. Both minerals can flood and dilute the heavy mineral assemblage and a method to remove the two phases was employed.
Baryte and to a lesser extent anhydrite grains were removed from heavy mineral separates through selective pulverisation using a standard dental amalgamator and acrylic balls. The vibration of the dental amalgamator and the acrylic balls pulverises softer minerals, leaving only more resistant minerals because of the low hardness of baryte and anhydrite (Mohs scale 3 to 3.5) compared to other heavy minerals. This method was adapted and modified from the procedure described by the Arizona Laserchron Center at the University of Arizona (listed as Processing Samples (Min Sep): WIG-L-BUG removal of softer materials) at: https:// sites. google. com/ laser chron. org/ arizo nalas erchr oncen ter/ home.
The heavy mineral separates and the acrylic balls of 1/8 mm size were placed into a stainless-steel capsule, which were securely placed onto the dental amalgamator for a short period of time (up to one minute). The vibration of the amalgamator destroyed the targeted soft minerals. The modified heavy mineral separates were rinsed using isopropanol through a 20 µm sieve. The isopropanol and pulverised baryte mixture created a ``milky'' mixture. The remaining hard minerals were washed with a small amount of acetone under a fume hood and subsequently examined under the microscope. If the resulting mixture was not milky or there were still large amounts of baryte visible under the microscope, the process was repeated. Heavy mineral separates were split before employing the method and analysed separately to assess the success of the method, but also to assess the effect on other heavy mineral grains.

LA-ICP-MS zircon geochronology
For zircon concentration the separates were further processed with a Frantz magnetic barrier separator and immersed into di-iodomethane at 3.3 g/cm 3 to maximise the purity of the zircon separates. Zircon grains were imaged in transmitted light to detect cracks or inclusions and cathodoluminescence imaging was performed to identify zoning prior selection of analysis spots for the laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS).
The methodology used follows the detrital zircon geochronology analysis scheme described in Breitfeld et al. (2020a). U-Pb LA-ICP-MS analysis was performed at Birkbeck College, University of London with a New Wave NWR 193 nm laser ablation system coupled to an Agilent 7700 quadrupole-based plasma ICP-MS with a two-cell sample chamber. A spot size of 25 μm was used for the ablation. The Plešovice zircon standard (337.13 ± 0.37 Ma; Sláma et al. 2008) and a NIST 612 silicate glass bead (Pearce et al. 1997) were used to correct for instrumental mass bias and depth-dependent inter-element fractionation of Pb, Th and U. Data reduction was achieved with the GLITTER software (Griffin et al. 2008). The data were corrected using the common lead correction method of Andersen (2002), which is used as a 204 Pb common lead-independent procedure. Supplementary Table 2 displays results of measured Plešovice zircons (weighted mean age: 337.11 ± 0.32 Ma, n = 159, MSWD = 0.08). IsoplotR (Vermeesch 2018) was used for calculation and illustration of zircon standard results (Supplementary Fig. 1).
For grains older than 1000 Ma the age was that obtained from the 207 Pb/ 206 Pb ratio. For ages younger than 1000 Ma, the ages from the 238 U/ 206 Pb ratio were used, because 207 Pb cannot be measured with sufficient precision, resulting in large analytical errors (Nemchin and Cawood 2005). Concordance was tested using a 10% threshold between the 207 Pb/ 206 Pb and 206 Pb/ 238 U ages for ages greater than 1 Ga and between the 207 Pb/ 235 U and 206 Pb/ 238 U ages for ages below 1 Ga. Age histograms and probability density plots were created using an R script that adopts the approach of Sircombe (2004) for calculating probability density. A few cores and rims were observed in CL images, and both sites were analysed following the approach by Zimmermann et al. (2018) to detect all age peaks, which is important for provenance studies. Analytical results are presented in Supplementary Table 3 for Cuu Long Basin and Supplementary Table 4 for Nam Con Son Basin. CL images with laser ablation spots of all zircons can be found in Supplementary  Fig. 2. For a simple illustration of the detrital zircon data we use a Multi-Dimensional Scaling (MDS) plot for detrital zircon age distributions (Vermeesch 2013). The MDS plot is based on a dissimilarity matrix between pairwise samples, using here the two-sample Kuiper distance test, and generated with the R provenance package by Vermeesch et al. (2016).

Biostratigraphy
The Cuu Long Basin samples are dominated by clastic sediments with strong fluvial or deltaic influence and only very rare non-age diagnostic foraminifera were observed. Samples from the Nam Con Son Basin yielded foraminifera which enable high resolution biostratigraphy for the Dua Formation (T30) and the Thong Formation (T40&50). The lowermost analysed succession, the Cau Formation, did not yield age-diagnostic foraminifera. The Dua Formation is dated as Early Miocene with the lower part having a foraminifera assemblage belonging to the Planktonic Foraminiferal zones N5 to N6 (presence of Catapsydrax stainforthii and Catapsydrax dissimilis) with an age of 21.0 to 17.2 Ma. Fragments of upper Oligocene to lower Miocene reefal larger benthic foraminifera (Lepidocyclina sp.) are reworked into this inner neritic foraminiferal assemblage. The uppermost sample (NCSa-2) belongs to zone N7 (based on the presence of Globorotalia miozea and Globigerina connecta) with an age of 17.2 to 17.0 Ma (Table 1). The overlying Thong Formation samples yielded a Langhian foraminifera assemblage diagnostic for Planktonic Foraminiferal zone N8b-N9a (presence of Praeorbulina glomerosa) with an age of 15.4 to 14.5 Ma (Table 1). This suggests a short hiatus between the two formations as zone N8a was not observed. Relative stratigraphic position of Nam Con Son Basin samples with age diagnostic foraminifera is displayed in Fig. 4. Assorted and index foraminifera photomicrographs from the Nam Con Son Basin samples are displayed in Fig. 5.

U-(Th)-Pb zircon geochronology
All of the samples analysed yielded zircons significantly older than the depositional ages, and therefore, no age estimate for the formations can be made from the detrital zircons.

Tra Tan Formation (D)
Sample CLB9.1 from the lower part of the Tra Tan Formation was analysed and 128 concordant U-Pb zircon ages were acquired from 134 zircons. CL imagery revealed predominantly oscillatory zoned or sector zoned euhedral zircons. No obvious core-rim structures were observed. A single grain had a core of 272 ± 5 Ma and with a slightly younger age of 257 ± 6 Ma from edge of the grain. The sample is dominated by Cretaceous zircons, which form about 87% (111 out of 128) of the zircon assemblage (Fig. 6). Cretaceous ages range from c. 70 to 122 Ma with a dominant narrow peak between 90 and 100 Ma. There are 16 ages that scatter from the Early Jurassic to the Devonian, with a small Permian population (11 ages). The oldest grain is 1758 ± 22 Ma and the youngest is 70 ± 1 Ma.

Lower Bach Ho Formation (BI.1)
Sample CLB6.1 from the Lower Bach Ho Formation was analysed and 135 detrital zircon grains yielded 131 concordant ages. Zircons are generally oscillatory zoned, and no core-rim structures were observed. The sample Table 1 Biostratigraphy results of the Cuu Long and Nam Con Son basins. Samples from well NCSa yielded a number of age diagnostic foraminifera. Age based on first appearance Planktonic Foraminiferal zones after BouDagher-Fadel (2015, 2018b (Fig. 6). The only other significant Precambrian age population lies between 700 and 800 Ma. The oldest analysed zircon is 3541 ± 5 Ma. The youngest three zircons are Oligocene to Eocene in age and range from 35.1 to 50 Ma, but do not overlap in error and are significantly older than the depositional age. Similar young zircons were not observed in the other Cuu Long Basin samples.

Nam Con Son Basin
Cau Formation (T20) The Cau Formation was analysed in three samples from cuttings (NCSa-7, NCSb-19, NCSb-23) and two samples from core (NCSc-1, NCSc-2). The combined age plot for the five samples is displayed in Fig. 7

Dua Formation (T30)
The Dua Formation was analysed in samples NCSa-4, NCSb-10, and NCSb-13. All three samples show very similar detrital zircon age populations and are displayed together as NCS-2. 398 concordant ages were acquired from 403 zircon grains. Zircon grains are relatively similar to the Cau Formation in being dominated by dark grey CL rounded to sub-rounded varieties. The Dua Formation has two main zircon age populations, a Permo-Triassic population that forms a double peak in the Late Permian and early Late Triassic and an older Paleoproterozoic age population around c. 1.8 to 1.9 Ga (Fig. 7). There are almost equal numbers of Phanerozoic (206) and Precambrian (192) zircons. Besides the Permo-Triassic age peak, other Phanerozoic grains are Cretaceous, Upper Jurassic and Ordovician-Silurian with some scattered ages between. The Precambrian has a very wide age population of c. 600 to 1150 Ma, a dominant peak at 1.8-1.9 Ga and a small peak at c. 2.5 Ga. The oldest age was 3693 ± 6 Ma in NCSa-4, and the youngest zircon was 41.9 ± 0.5 Ma in NCSb-10.

Thong Formation (T40&50)
The Thong Formation was analysed in samples NCSa-1 and NCSb-1. Both samples show very similar zircon age populations and are displayed together as NCS-1. 164 concordant ages were acquired from 169 zircon grains. Zircons show predominantly homogenous or oscillatory zoning in CL imagery for euhedral grains, and dark CL for subrounded to rounded grains that dominate the assemblage. Dark areas were usually associated with Precambrian ages. Like the underlying Dua Formation, the Thong Formation is dominated by Triassic zircon ages with a significant number of Precambrian zircons (Fig. 7). In contrast to the Dua Formation, the number of Permian and Cretaceous zircons is lower.

Heavy mineral analysis
Heavy mineral results in this study were acquired using three different methods, which in general yielded very comparable results. EDS-SEM and QEMSCAN are electron beam methods which cannot differentiate between TiO 2 polymorphs, whereas Raman spectroscopic analysis can differentiate between rutile, brookite and anatase. Raman results generally indicated slightly more total TiO 2 polymorphs than the electron beam methods, which is likely a function of composite anatase grains that are below the threshold for identification by the electron beam methods, but still result in a strong anatase Raman shift signal. The proportion of TiO 2 polymorphs is likely, therefore, slightly overestimated in the samples analysed by Raman spectroscopy compared to the EDS or QEMSCAN results. Nevertheless, results are consistent. For simplification, in this paper the three polymorphs rutile, brookite and anatase are listed as rutile.
Other polymorphs (such as aluminosilicates) were not found. Baryte was discarded from the heavy mineral assemblage, because it was used as a drilling additive. Opaque minerals were also analysed, but due to identification uncertainties using different methods as well as a possible diagenetic origin (e.g., pyrite) no provenance interpretation is given. Heavy mineral proportions are listed in Table 2 and count numbers are listed in Supplementary Table 5.

Cuu Long Basin
The heavy mineral assemblage of the Tran Tra Formation (D) sample CLB9.1 consists of abundant translucent heavy minerals which make about 87% of the total assemblage. Opaque minerals are subordinate, with pyrite (12.5% of the total heavy mineral assemblage) being the most abundant. The translucent heavy mineral assemblage is dominated by apatite (39.7%), rutile (26%) and zircon (22.7%) (Fig. 8A).
The heavy mineral assemblage of the overlying Lower Bach Ho Formation sample (BI.1) CLB6.1 is also dominated by translucent heavy minerals which make c. 92% of the  (Fig. 8A). Fluorite, monazite, sphalerite and topaz are present in traces. The Con Son Formation (BII) heavy mineral samples CLB2.1, CLB2.3 and the Dong Nai Formation (BIII) sample CLB1.1 show very similar heavy mineral proportions and record a major change from the underlying Bach Ho Formation. In contrast to the underlying two formations the heavy Fig. 7 Detrital zircon age histograms with probability density curves for the Nam Con Son Basin samples in stratigraphic order. Samples from the respective formations have been combined (X = number of samples). The Cau and Dua Formation samples display a heterogeneous zircon age distribution with characteristic peaks in the Permian-Triassic, Cretaceous, Ordovician-Silurian and Paleoproterozoic (c. 1.8 Ga). The Dua Formation has thereby significant more Precambrian ages as the Cau Formation, which may be related to unroof-ing of a source area. The Cau Unconformity separates the two formations. The Thong Formation above the Dua Formation shows an increase in 800 to 1000 Ma and c. 2.5 Ga zircons, typically for Sibumasu, suggesting a new river system supplying material. The Dua Unconformity marks the boundary at c. 15.4 to 17 Ma. Bin size of 10 Ma for Phanerozoic ages and 50 Ma for Precambrian ages. Supplementary Figure 3 displays the histograms for the individual NCSB samples Table 2 Heavy mineral proportions of the analysed CLB and NCSB samples. Mineral ratios after Morton and Hallsworth (1994), Morton (2007), and ZTR (zircon-tourmaline-rutile) index after Hubert (1962)   Pyrite makes up to 50% of the total heavy mineral assemblage and hematite (up to 21%) and ilmenite (up to c. 7%). Translucent heavy minerals form only around 40 to 54% of the total heavy mineral assemblage with rutile (up to 48.6%), apatite (9.7 to 21.1%), tourmaline (15.3 to 18.3%) and garnet (11.7 to 16.1%) being the most abundant (Fig. 8A). Zircon (up to 5.6%), monazite (up to 3.3%), pyroxene (up to 4.2%), epidote (up to 2.8%), amphibole (up to 2.3%) and Cr-spinel (up to 1.4%) are subordinate (Fig. 8A). Traces of xenotime, sphalerite, titanite and staurolite were also observed. The Tra Tan Formation and the Bach Ho Formation samples show some similarities in relevant heavy mineral indices, despite the high amount of epidote in the Bach Ho Formation. The ZTi and CZi values and the moderate RZi are similar in the lower Bach Ho Formation and the Tra Tan Formation samples (Fig. 9A). The two samples differ in ZTR which decreases sharply in the Lower Bach Ho Formation as a result of abundant epidote, in GZi which increases as garnet increases and in ATi which decreases sharply as apatite decrease (Fig. 9A), thus indicating some source differences for the two formations. The Upper Bach Ho Formation sample (CLB4.1), despite its very similar heavy mineral proportions to the Lower Bach Ho sample (CLB6.1), has indices that indicate a change within the formation. Like the Lower Bach Ho Formation ZTR is low, ATi is moderate, and CZi is 0 (Fig. 9A). In contrast, GZi and RZi increase significantly and ZTi decreases, which is more similar to the overlying formations (Fig. 9A). Up section the heavy mineral indices show a major change between the Bach Ho Formation and the overlying Con Son and Dong Nai Formations. The ZTR values of the Con Son and Dong Nai Formation increases slightly from the underlying Bach Ho Formation as a result of the higher abundance of rutile and tourmaline (Fig. 9A). There is also an increase in RZi (up to 93) and a decrease of ZTi (down to 16) in the Upper Bach Ho Formation (Fig. 9A). In addition, there is a sharp increase in CZi (Fig. 9A), all supporting a significant source change at the base of the Con Son Formation, which may have started within the Upper Bach Ho Formation. The Dong Nai Formation shows a relatively similar heavy mineral assemblage with similar heavy mineral indices to the underlying Con Son Formation (Fig. 9A), suggesting no major source changes for the two formations. The ZTR-GZi bi-plot best displays the two different main assemblages observed in the CLB samples (Fig. 9C).
Translucent heavy minerals from the Nam Con Son Basin are generally dominated by ultra-stable minerals, indicated by the high ZTR values (up to c. 98%, Fig. 9B). The youngest analysed unit, the Thong Formation, has the lowest ZTR values (c. 65 to 75%, Fig. 9B) due to the presence of less

Con Son Dong Nai
Bach Ho Tra Tan Fig. 9 Heavy mineral indices plots, using the ZTR (Hubert 1962), RZi, GZi, ATi, ZTi and CZi (Morton and Hallsworth 1994;Morton 2007). A) CLB heavy mineral data shows a significant change at the Bach Ho Unconformity between the underlying Bach Ho Formation and the overlying Con Son. Indices ZTR, CZi, GZi and RZi best illustrate the change in provenance, while other indices are more variable. B) NCSB heavy mineral data is displayed per sample (left panel) and per seismic sequence/formation (right panel). Indices are very simi-lar, only the ZTR, ZTi and GZi show some variation from T20/30 to T40&50. C) ZTR-GZi biplots for the CLB and NCSB can be used to distinguish between the different formations. CLB data clearly separates into the Tra Tan-Bach Ho trend (red) and into the Con Son-Dong Nai (green). The three formation of the NCSB are also separated by the bi-plot with the Cau and Dua formations being very similar, and the Thong Formation having a slightly different heavy mineral assemblage stable heavy minerals, such as apatite and garnet (Fig. 8b). The other heavy mineral indices are very variable for the Nam Con Son Basin and are best displayed for formations and not individual samples (Fig. 9B) so that trends can be better established. The ATi ratio decreases from the Cau (T20) to the Dua Formation (T30), before increasing in the Thong Formation (T40&50) (Fig. 9B). At the same time the ZTi remains relatively stable and high in the Cau and Dua Formations, before sharply dropping at the top Dua sample and remaining low in the Thong Formation (Fig. 9B).
A higher input of garnet in the Thong Formation is indicated by the GZi values of c. 30 (Fig. 9B), higher than in the underlying Dua Formation. The Cau and Dua Formation samples have also ZTi values generally higher than in the overlying Thong Formation. Using ZTR and GZi indices, the three formations are relatively well separated (Fig. 9C).

Effect of baryte removal
Samples CLB2.1, CLB2.3 and NCSb-19 were subjected to the baryte removal technique (Supplementary Table 6). CLB2.1 was analysed in two separates with QEMSCAN. One separate was untreated and the other was treated to remove baryte. 540 translucent heavy mineral grains (excluding carbonates) were identified in the untreated sample with baryte forming 53.7% of the assemblage (Fig. 10A). The treated sample yielded 799 translucent heavy mineral grains with the proportion of baryte decreasing to 20.2% (Fig. 10A). The results from the two separates, after excluding baryte, are almost identical (Fig. 10A) and indicate that the baryte removal technique had no effect on the abundance of other translucent heavy minerals. With the decrease of baryte the count numbers were significantly increased. Anhydrite was not separately counted with QEMSCAN and is listed under other phases in the Supplementary Table 5. Sample NCSb-19 was also analysed with QEMSCAN in two separates. The untreated sample yielded only 45 translucent heavy minerals with baryte forming 60% of the assemblage (Fig. 10B). The treated sample yielded 559 translucent heavy mineral grains with the proportion of baryte decreasing to 38.5% (Fig. 10B). No apatite and tourmaline were found, reflecting the low number of translucent heavy mineral grains and the high proportion of baryte in the untreated separate, but proportions of other translucent heavy minerals are comparable. The analysis of CLB2.1 and NCSb-19 indicates the importance of baryte removal which results in an increase in the number of other heavy mineral grains and decreases the risk of missing low abundance minerals.
CLB2.3 was analysed with Raman spectroscopy, which also recorded the counts of anhydrite. Seventeen translucent heavy mineral grains were identified (excluding carbonates) from the untreated separate. Baryte and anhydrite each form 17.7% of the assemblage (Fig. 10C)  Anhydrite proportions increase in the treated separates, which indicates that further processing and testing is necessary yielded significantly more translucent heavy mineral grains (51 excluding carbonates). Baryte was not found and, therefore, had been completely removed from the sample, while anhydrite forms 19.6% of the assemblage (Fig. 10C). It is likely that anhydrite was broken down into smaller pieces, but still larger than the 20 µm sieve size. Another treated separate from CLB2.3 was analysed with SEM-EDS, which showed no baryte but abundant anhydrite (Fig. 10C), and suggesting breakdown of larger anhydrite grains into numerous > 20 µm fragments. It is, therefore, concluded that the short vibration time intervals removed the baryte, but for anhydrite removal a longer time interval is required, which requires more testing.
The tested samples showed no significant variations of heavy mineral abundances in treated and untreated samples, indicating that only baryte was removed. Count numbers were increased in the treated samples and identification of some less abundant heavy minerals was possible. For anhydrite contamination, the method is promising, but more tests with different settings need to be conducted.

Cuu Long Basin
The sediments analysed from the Cuu Long Basin were sourced by two different river systems. The first river, which we term the proto-Mekong 1, active from the Oligocene to Early Miocene, deposited the Tra Tan (seismic group D) and Bach Ho (seismic group BI) Formations. These formations have Cretaceous-dominated detrital zircon age populations (Fig. 6), which indicate a proximal Da Lat Zone basement source (Nguyen et al. 2004;Shellnutt et al. 2013;Hennig-Breitfeld et al. 2021). The translucent heavy mineral assemblages vary in the three analysed samples, but relatively low ZTR values are characteristic, with either high proportions of epidote or apatite (Fig. 8A). The zircons are almost entirely Cretaceous, very similar to the onshore Oligocene to Miocene Di Linh Formation ( Fig. 11; ). The heavy mineral assemblage of the Di Linh Formation differs from the offshore samples in being rich in ultra-stable grains , which might indicate hydraulic sorting or mineral breakdown caused by acid groundwaters on land. Epidote, which is dominant in the Bach Ho samples, was also reported from the Di Linh Formation , whereas all other onshore and Cuu Long Basin formations are almost barren of epidote. The lowermost analysed Cuu Long Basin sample CLB9.1 of the Tra Tran Formation has no epidote in contrast to the overlying Bach Ho Formation samples. This implies a fresh epidote-free granitoid source for the older Tra Tran Formation (seismic group D), while the epidote in the Bach Ho Formation samples was derived from altered plagioclase and amphibole in deeper altered granitoids. Hennig-Breitfeld et al. (2021) reported epidote alteration in most onshore granitoids and in associated volcanic rocks, and Jagodziński et al. (2020) reported elevated epidote concentrations in the eastern Mekong River distributaries in the Da Lat Zone and east of the Mekong River Delta from recent deposits, indicating that the Da Lat Zone is a source for epidote.
The overlying Con Son Formation records a major provenance change suggesting input from a different Mekong river in the Middle Miocene. Zircon age populations are dominated by Triassic ages with a few Precambrian grains of c. 800 Ma, c. 1.8 Ga, and 2.5 Ga, as well as Cretaceous and Ordovician-Silurian grains (Fig. 6). The detrital zircon ages are typical of Indochina with a dominance of Triassic Sukhothai arc ages (Sevastjanova et al. 2011;Burrett et al. 2014), but could also indicate a far larger drainage area including the Qiangtang, Yangtze or Songpan-Garze blocks, where similar age spectra are known (Wang et al. 2020 and references therein). Rutile, tourmaline and garnet proportions increase significantly in the Con Son Formation (Fig. 8A), and high proportions of these ultra-stable heavy minerals indicates multi-recycling of metamorphic sources. Mineral ratios GZi, CZi, ZTR also indicate a major source change (Fig. 9A, C). It is unclear exactly when the Con Son Formation was deposited unconformably above the Bach Ho Formation and precisely when this significant source change occurred. Morley et al. (2019) interpret the top of the Bach Ho Formation as c. 20.5 Ma which could be the age of the change. However, the unconformity might have cut down into the underlying Bach Ho Formation during a hiatus in deposition. The rearrangement of the river systems would suggest some period of non-deposition/erosion. Hiep (2017) suggested a Middle Miocene age (starting at c. 16 Ma) for the beginning of deposition of the Con Son Formation. The major unconformity we recognise here, the Bach Ho Unconformity, may, therefore, have an age between 20.5 and 16 Ma. Heavy minerals of the Dong Nai and Con Son formations are very similar, with most heavy mineral indices showing comparable values, but there is also a minor change in zircon ages from the Con Son Formation to the Dong Nai Formation (Fig. 6). Although the provenance remained the same, suggesting the same river system, there is an Eocene zircon population in the Dong Nai Formation indicating an additional source which was not available for the underlying Con Son Formation. The Eocene zircons were probably derived from magmatic rocks in the Da Lat Zone or southern Indochina that were unroofed during the Late Miocene. Fyhn et al. (2010) suggested Palaeocene to Early Eocene inversion in the Phuquoc-Kampot Som Basin, which if associated with magmatism, could be a potential proximal source.

Nam Con Son Basin
Three formations were analysed from the Nam Con Son Basin. The Cau (T20) and Dua (T30) formations have very similar heavy mineral assemblages consisting mainly of rutile with zircon, tourmaline, monazite, apatite, and other less abundant minerals, such as garnet or Cr-spinel ( Fig. 8B). An assemblage with high ZTR values is usually typical of multi-recycled sedimentary sources. The detrital zircon age populations in both formations are dominated by Permian-Triassic and Paleoproterozoic ages around 1.8-1.9 Ga (Fig. 7). In addition, there are significant Cretaceous, Jurassic, Silurian, Neoproterozoic, and Proterozoic-Archean (c. 2.5 Ga) age populations. The detrital zircon ages suggest similar sources in Indochina, possibly as far as the Qiangtang block. The Cau Formation contains almost twice the number of Phanerozoic detrital zircons compared to Precambrian grains, while the ratio decreases in the Dua Formation to almost equal proportions, and this can be used to distinguish the two formations (Fig. 7). The unconformity between the two formations is here referred to as the Cau Unconformity.
Heavy mineral assemblages from the overlying Thong Formation (T40&50) record another source change. There is a decrease in rutile and an increase of apatite and tourmaline (Fig. 8B). The assemblage is still dominated by ZTR minerals and suggests multi-recycled older sedimentary rocks as source. The detrital zircon age populations are different from the underlying formations (Fig. 7). The dominant age population is Triassic, with only a small population of Permian zircons. The Cretaceous population, previously very prominent, decreases significantly and indicates the Da Lat Zone was no longer an important source. Similarly, the Paleoproterozoic (c. 1.8 to 1.9 Ga) population decreases. Neoproterozoic (especially between 850 to 1000 Ma and at c. 500 Ma) and Proterozoic-Archean (c. 2.5 Ga) zircons increase in the Thong Formation. The Neoproterozoic ages are typical of the Sibumasu terrane of the Malay Peninsula (Hall and Sevastjanova 2012;Lin et al. 2013), especially without the 1.8 Ga population typical of Indochina. Samples from Sibumasu (Hall and Sevastjanova 2012) are also dominated by Triassic detrital zircons with few Permian grains. This suggests that the Thong Formation was directly sourced from the south, where the drainage area included parts of East Malaya and Sibumasu. The Thong Formation is part of the syn-rift 2 phase of the Nam Con Son Basin. The unconformity recognised between the Dua and Thong formations is here named Dua Unconformity. The age of the unconformity is relatively well established, at about 16 Ma, since the underlying Dua Formation was dated in this study as young as 17 Ma and the Thong Formation may be as old as 15.4 Ma (Table 1).

Mekong River evolution
The age of formation of the Mekong River, the source areas of sediments, and the river course through time continue to be debated. Wang et al. (2020) for example used the term paleo-Mekong for an Early Cretaceous river that may have drained the Qiangtang block in a similar way to the presentday Mekong River. However, it is difficult to link the present Mekong River to any Early Cretaceous river in Indochina or South China behind an active Andean-type margin to the east (Shellnutt et al. 2013;Xu et al. 2016;Breitfeld et al. 2017Breitfeld et al. , 2020bHennig et al. 2017aHennig et al. , 2017bHennig-Breitfeld et al. 2021) when the topography was probably very different. Most researchers discuss the Mekong River evolution from after the Cenozoic India-Asia collision (e.g., Hutchison 1989;Clark et al. 2004;Clift et al. 2006;Zhang et al. 2019), and we focus here on the Oligocene-Miocene to recent development of the Mekong. Figure 12 illustrates our interpretation of the evolution of the Mekong River using our new provenance data and guided by the present-day major river courses of Indochina and Vietnam. The proto-Mekong 1 (Fig. 12A), of Oligocene to Early Miocene age, is the Phase 1 proto-Mekong named by  with a delta slightly north of the present-day Mekong River and a tributary flowing from the Da Lat Zone inferred from the Cretaceous-dominated zircon populations of the Tra Tan, Bach Ho and Di Linh formations (Fig. 11), and the immature heavy mineral assemblages of the former two. We suggest that at this time there was no major Mekong River system draining from far to the north in the India-Asia collision zone that supplied sediment to the offshore Vietnam Cuu Long Basin, instead the proto-Mekong 1 with a restricted drainage provided sediment. This is different to some other authors (e.g., Hutchison 1989;Zhang et al. 2019) who have argued a proto-or paleo-Mekong was in place in the Oligo-Miocene draining the Tibetan Plateau. Hutchison (1989) favoured a straight proto-Mekong west of the Khorat Plateau that flowed along the route of the present Chao Phraya River into the Gulf of Thailand before Late Cenozoic faulting changed its course. Brookfield (1998) argued that the proto-Mekong flowed via the Yom River towards the Chao Phraya until river capture in the Late Cenozoic resulted in the present-day Mekong course. Clark et al. (2004) proposed that the Upper Mekong flowed into a paleo-Red River before capture formed the modern Mekong River. None of these rivers with headwaters in the Himalaya or Tibetan Plateau had river courses or delta locations similar to the present-day Mekong River. Clift et al. (2008) modelled several unnamed rivers draining the Tibetan Plateau flowing to the south, and like , interpreted an early proto-Mekong with drainage restricted to Vietnam. The various proposed courses of the paleo-or proto-Mekong Rivers are discussed in detail in Carling (2009).
At 25 Ma (Fig. 12A) we propose using the term proto-Chao Phraya for the river system identified by Hutchison (1989) and Brookfield (1998) that drained southwards from the India-Asia collision zone towards the Gulf of Thailand/ Malay Basin. This river system likely carried sediment to the Nam Con Son Basin. We suggest there were two different strands of the proto-Chao Phraya, a western strand that would become the Salween and an eastern strand which would become the proto-Mekong 2 from the Middle Miocene onwards. This proto-Mekong 2 was formed by capture of the eastern strand of the proto-Chao Phraya by the proto-Mekong 1 (Fig. 12B). The new river had a course similar to the present-day Mekong and drained a large Indochina to Himalayan region, consistent with Nie et al. (2018) and Zhang et al. (2019) who also suggested a large proto-Mekong River draining back into the Tibetan Plateau from at least the Mid-Miocene. Detrital zircon age populations from the Con Son and Dong Nai formations (Cuu Long Basin) are very similar to the present-day Mekong zircon signatures ( Fig. 13), indicating a similar drainage to that of today. Its delta was likely still in the vicinity of the Da Lat Zone as the present-day Mekong River delta was not formed until the Late Pleistocene to Holocene (e.g. Ta et al. 2002Ta et al. , 2021Tanabe et al. 2003;Nguyen et al. 2010;Xue et al. 2010;Dung et al. 2013;Thanh et al. 2021). The western strand of the proto-Chao Phraya may have developed into the Salween River in the Middle Miocene leaving a reduced Chao Phraya   (Fig. 12B). The Irrawaddy River (or Ayeyarwady) of Central Myanmar west of the Salween has been in its current geographical position since the late Oligocene, c. 27 Ma (Gough et al. 2020) draining the Indo-Myanmar Ranges in the west and the Eastern Himalayan Syntaxis in the north. The northeast tributary of the proto-Mekong 2 was to become the Da Lat River (Phases 2 and 3 of the proto-Mekong of  in the Plio-Pleistocene (Fig. 12C) (Mitchell 1986;Racey et al. 1997;Morley 2012) or Late Cenozoic to Pleistocene basalt magmatism associated with uplift in the Da Lat Zone (Rangin et al. 1995;Hoang and Flower 1998;Lee et al. 1998;An et al. 2017) seem to have a major influence at least on the course of the lower Mekong (Fig. 12D). The fill of the Nam Con Son Basin was not directly linked to the proto-Mekong system. Detrital zircon ages of the Oligocene to Early Miocene Nam Con Son Basin sediments are very similar to the Middle to Late Miocene fills of the Cuu Long Basin (Fig. 13). This suggests that the proto-Chao Phraya (Fig. 12A) which drained from far to the north into the Tibet Plateau fed sediment into the Nam Con Son Basin before the Middle Miocene via the then fluvial area of the Gulf of Thailand (Fig. 12A). After the Dua Unconformity in the Middle Miocene the Nam Con Son Basin was filled mainly by the Kelantan River (Fig. 12B), which drained back into Sibumasu. The interpretation that the Pleistocene fill of the Nam Con Son Basin was derived from a prograding paleo-Mekong (e.g. Matthews et al. 1997), therefore, seems unlikely. Figure 14A illustrates the paleogeography of the southern South China Sea region in the late Oligocene before the Bach Ho and Dua unconformities, and Fig. 14B after the unconformities in the Middle Miocene with the major river courses. The results of this study support the interpretation of  of a major mid-Miocene unconformity in southern Vietnam and Indochina at which the drainage network was reorganised.

Detrital zircon age comparison of the proto-Mekong(s) with the proto-Chao Phraya
The MDS plot (Fig. 15) of the detrital zircon age data from this study compared to the present-day Mekong data (Bodet and Schärer 2000;Clift et al. 2006;Nguyen et al. 2018) and the onshore Da Lat Zone sedimentary rocks , shows the evolution of the different systems and their similarities by scaling the data of samples/formations into two dimensions (Vermeesch 2013) for a faster visual representation. The Oligocene to Early Miocene fill of the Cuu Long Basin (Tra Tan, Bach Ho formations) was derived from the proto-Mekong 1, which had a localised drainage of the Da Lat Zone and a river delta along the SE Vietnam coastline directly adjacent to the Cuu Long Basin. The onshore Di Linh Formation represents the last onshore remains of this system. This fill of the Cuu Long Basin is estimated to be as thick as c. 4000 m (Cuong and Warren 2009) and indicates that despite the restricted drainage, large amounts of material were eroded and deposited by the proto-Mekong 1. In contrast, the Middle to Late Miocene Cuu Long Basin sediments (Con Son and Dong Nai formations) resemble the present-day Mekong deposits in their detrital zircon age populations and were part of the proto-Mekong 2 system (Figs. 13, 15). The Cretaceous population in both formations, and the Jurassic zircon population in the Con Son Formation indicate a slightly different drainage compared to the present-day, probably involving some minor input from the Da Lat Zone that is by-passed by the present-day Mekong (Fig. 12B). The Cretaceous-dominated proto-Mekong 1 age distribution signature is illustrated in Fig. 11. Noticeable in the histograms is that the offshore formations have a wide Late Cretaceous age peak from c. 80 to 110 Ma, while the Di Linh Formation has a bimodal Cretaceous age distribution, suggesting local igneous bodies had strong influence on the age distribution. The offshore samples have also a slight population in the Permian-Triassic, which is almost absent in the Di Linh sample that suggests input of older basement in the vicinity of the delta (e.g., Chau Thoi). The Da Lat River developed its drainage towards the north in the Pliocene, resulting in an increase of Triassic and Precambrian zircons derived from Indochina and South China basement ( Fig. 11; ). The Ba Mieu and Song Luy formations deposited by the Da Lat River (Phase 2 and 3 of the proto-Mekong in ) are clearly distinguishable from the proto-Mekong 1 and the proto-Mekong 2/proto-Chao Phraya by their detrital zircon ages (Figs. 11 and 15).
The lower Nam Con Son Basin successions (Cau/T20 and Dua formations/T30) are also very similar to the upper Cuu Long Basin samples (Fig. 15), especially the Cau Formation, indicating a similar provenance and supporting the suggestion that the headwaters of the proto-Chao Phraya would later become the proto-Mekong 2. The heavy mineral assemblages are mineralogically comparable, but mineral proportions in the Cuu Long Basin and Nam Con Son Basin samples vary, possibly as a result of sorting. Garnet is relatively abundant in the Cuu Long Basin but present only in low amounts in the Cau and Dua formations of the Nam Con Son Basin. The Cretaceous zircons in the Cau and Dua formations (Figs. 7 and 13) were likely sourced by highs at the basin margin (e.g. Con Son Swell) rather than input from the Da Lat Zone. The overlying Thong Formation (T40&50) has small differences in detrital zircon ages that is represented by being slightly off T20 and T30 (Fig. 15), and has also a slightly different heavy mineral assemblage (Fig. 8B). The increase in Cambrian and Neoproterozoic zircons (Figs. 7 and 13) is a typical Sibumasu signature, indicating a southwestern source in the Malay Peninsula drained by the Kelantan River. A similar drainage for the Nam Con Son Basin fill was concluded by Morley et al. (2011) and Morley and Morley (2013).

Conclusions
Detrital zircon ages and heavy mineral assemblages of Cenozoic sedimentary formations in on-and offshore SE Vietnam can be used to reconstruct the drainage history of major river systems in Indochina. The Oligocene to Early Miocene basin fill of the Cuu Long Basin (Tra Tan and Bach Ho formations) was derived almost entirely from uplifted Dal Lat Zone basement highs along the basin margin by the proto-Mekong 1. The Bach Ho Unconformity marks a major source change and the overlying Con Son and Dong Nai formations have a provenance similar to the presentday Mekong, indicating that by the Early-Middle Miocene boundary the proto-Mekong 2 had a course similar to the present-day. The proto-Mekong 1 captured the headwaters of the proto-Chao Phraya to form the proto-Mekong 2. The Oligocene to Early Miocene basin fill of the Nam Con Son Basin (Cau and Dua formations) has detrital zircon ages that resemble the present-day Mekong and the upper Cuu Long Basin fill, suggesting a similar provenance. We propose that in the Oligocene to Early Miocene a proto-Chao Phraya, with a drainage area into the Himalaya highlands, filled the Nam Con Son Basin. The Thong Formation of the Nam Con Son Basin, unconformable on top of the Dua Formation, has a partly Sibumasu provenance and suggests input from the Malay-Thai Peninsula. The proto-Mekong 1 and 2 did not  Hennig-Breitfeld et al. 2019;Breitfeld et al. 2020a). A) Reconstruction at c. 25 Ma before the Bach Ho and Dua unconformities. The CLB is filled by the proto-Mekong 1 with a drainage area of the Da Lat Zone. The proto-Chao Phraya with the Kelantan filled the NCSB with the Gulf of Thailand being a fluvial-lacustrine area. The Sunda River drains the Sunda Shelf and had its delta in the NW Borneo region (based on Hennig-Breitfeld et al. 2019;Breitfeld et al. 2020a). B) Reconstruction at c. 15 Ma after the major drainage reorganisation. Proto-Mekong 1 has captured headwaters of proto-Chao Phraya to form the proto-Mekong 2 with a similar provenance and river course as the present-day that fills the CLB. The Kelantan fills the NCSB from the Malay-Thai Peninsula. The Da Lat River in SE Vietnam may begin to develop from the northern strand of the former proto-Mekong 1. The headwaters of the proto-Chao Phraya were captured, leaving a Chao Phraya with restricted drainage similar to the present-day provide sediment to the Nam Con Son Basin. In both basins unconformities around the Early to Miocene boundary (20.5 to 16 Ma for the Cuu Long Basin Bach Ho Unconformity and 17 to 15.4 Ma for the Nam Con Son Basin Dua Unconformity) are marked by a major source change associated with a drainage reorganisation.
Acknowledgements The project was funded by the SE Asia Research Group, which is sponsored by a consortium of industry bodies. Talisman Vietnam (now Repsol Vietnam) is thanked for financial support, sample shipment and geological background. Tim Pearce of Chemostrat Ltd. is thanked for support on a number of additional heavy mineral Raman spectroscopy analyses. We thank Martin Rittner and Andy Carter for help with the LA-ICP-MS analysis at UCL/Birkbeck College (UK). Jenny Omma of Rocktype is kindly acknowledged for the QEMSCAN results. Many thanks also to the South Vietnam Geological Mapping Division, in particular Dinh Quang Sang, Mai Kim Vinh and Trinh Van Long, for the insightful discussions in the field and help with the sample logistics. Ulrich Riller is thanked for editorial assistance, and topic editor Nora Noffke, Clive Burrett and an anonymous reviewer are thanked for comments on the manuscript.
Funding Open Access funding enabled and organized by Projekt DEAL.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.  (Clift et al. 2006); X, Y and W (Nguyen et al. 2018). Red-shaded area symbolises the Oligocene to Lower Miocene proto-Mekong 1 (D/BI Tra Tan and Bach Ho of CLB, and onshore Di Linh Formation). Yellow-shaded area represents the Plio-Pleistocene Da Lat River (onshore Song Luy and Ba Mieu formations). The green-shaded area represents the proto-Chao Phraya before (NCSB T20/30 Cau and Dua) and the proto-Mekong 2 after capture (CLB BII/BIII Con Son and Dong Nai) by the proto-Mekong 1, as well as the Kelantan signal (NCSB T40&50 Thong). The purple area represents the present-day Mekong. Arrow 1: represents the evolution from proto-Mekong 1 to Da Lat River (Phase 2 and 3 proto-Mekong in , arrow 2: represents the capture of the proto-Chao Phraya headwaters by proto-Mekong 1 to form the proto-Mekong 2, and 3: the change from Indochina sources for the T20/30 to Sibumasu source for T40&50 in the NCSB. The MDS plots are not sufficient to distinguish clearly between the Sibumasu-Indochina (Thong Formation) and the Indochina-Qiangtang provenance (Cau, Dua, Con Son, Dong Nai formations, present-day Mekong). Supplementary Figure 4 illustrates the same plot with NCSB samples separately plotted. Lines represent nearest neighbour. Source data for the MDS plot can be found in Supplementary Table 7